milvus

**Milvus** is an **open-source vector database built for scalable AI similarity search** — designed to handle billions of vectors with millisecond query latency, supporting multiple index types (IVF, HNSW, DiskANN) and hybrid search combining vector similarity with scalar filtering. **Key Features** - **Scale**: Handles 1B+ vectors with distributed architecture. - **Index Types**: IVF_FLAT, IVF_SQ8, HNSW, DiskANN for different speed/accuracy tradeoffs. - **Hybrid Search**: Combine vector similarity with attribute filtering. - **Cloud**: Zilliz Cloud for managed deployment. - **GPU Acceleration**: NVIDIA GPU-powered indexing and search. **Use Cases**: RAG retrieval, recommendation systems, image similarity, anomaly detection, drug discovery. **Comparison** - vs Pinecone: Open-source, self-hosted option, more index flexibility. - vs Qdrant: Better GPU support, more mature at billion-scale. - vs FAISS: Full database features (CRUD, filtering) vs library-only. Milvus is **the production choice for billion-scale vector search** — combining open-source flexibility with enterprise-grade scalability.